Advancing short-term prediction of suicidal behavior among high-risk youth: Real-time monitoring of suicide risk after psychiatric emergency department visits

Suicide, the second leading cause of death among youth in the United states, is a major public health crisis. Although distal risk factors for suicidal behavior are well-established, previous research—which has largely relied on cross-sectional or longitudinal studies spanning over long assessment intervals—has revealed relatively little about short-term predictors of suicide risk. This is a significant gap because prediction of short-term risk is critically important for clinical providers who are tasked with recognizing youth at imminent suicide risk to make disposition decisions to prevent a suicide attempt or suicide. Improving prediction of imminent suicide risk has also been highlighted as a priority by the National Action Alliance for Suicide Prevention. Advancements in technology that allow for real-time assessment of theoretically- and clinically-informed suicide risk indicators—such as via ecological momentary assessment (EMA), passive mobile sensing, and wearable physiological monitoring devices (wearable sensors)—hold immense potential to identify short-term suicide risk in the "real world." However, evidence regarding the validity and utility of this kind of data in predicting suicide risk still needs to be established. Specifically, translating EMA and passive mobile and sensor data into actionable predictors requires rigorous empirical validation before such data can be applied to inform clinical decision-making. Once validated, reliable markers of short-term suicide risk have the potential to inform the development of real-time personalized interventions delivered in individuals' natural environment when they are most needed. Thus, as a key step toward developing validated predictors of short-term suicide risk integrating subjective and passive objective data, we propose to develop and obtain initial predictive utility data for a multi-modal assessment paradigm designed to facilitate detection of short-term suicide risk in youth. We consider this approach among youth discharged from a psychiatric emergency department (ED), many of whom are vulnerable to experiencing post-discharge suicide-related crises (suicide attempts, return ED visits for suicide related concerns). This study's aims are to: (1) finalize a suite of predictors and measurements that will form a multi-modal assessment paradigm (EMA, passive mobile, and passive wearable sensor data) for detection of short-term suicide risk; and (2) examine its feasibility and initial predictive utility in predicting short-term suicide risk (suicidal behavior [actual attempts, interrupted attempts, and aborted attempts]; return ED visit for suicide-related reasons) among 75 youth (ages 18-24). EMA, mobile, and wearable sensor data will be collected from youth for two months following ED discharge. At the conclusion of this real-time monitoring phase of the study, participants will complete a semi-structured phone interview. Findings from this study will generate models predicting short-term suicide risk that will then be evaluated in a larger validation study. Thus, this study represents an important step toward developing robust and scalable methods for predicting proximal suicide risk in youth that, in turn, can inform real-time personalized interventions.